Discovery and mode of action of a novel analgesic β-toxin from the African spider Ceratogyrus darlingi

Spider venoms are rich sources of peptidic ion channel modulators with important therapeutical potential. We screened a panel of 60 spider venoms to find modulators of ion channels involved in pain transmission. We isolated, synthesized and pharmacologically characterized Cd1a, a novel peptide from the venom of the spider Ceratogyrus darlingi. Cd1a reversibly paralysed sheep blowflies (PD50 of 1318 pmol/g) and inhibited human Cav2.2 (IC50 2.6 μM) but not Cav1.3 or Cav3.1 (IC50 > 30 μM) in fluorimetric assays. In patch-clamp electrophysiological assays Cd1a inhibited rat Cav2.2 with similar potency (IC50 3 μM) without influencing the voltage dependence of Cav2.2 activation gating, suggesting that Cd1a doesn’t act on Cav2.2 as a classical gating modifier toxin. The Cd1a binding site on Cav2.2 did not overlap with that of the pore blocker ω-conotoxin GVIA, but its activity at Cav2.2-mutant indicated that Cd1a shares some molecular determinants with GVIA and MVIIA, localized near the pore region. Cd1a also inhibited human Nav1.1–1.2 and Nav1.7–1.8 (IC50 0.1–6.9 μM) but not Nav1.3–1.6 (IC50 > 30 μM) in fluorimetric assays. In patch-clamp assays, Cd1a strongly inhibited human Nav1.7 (IC50 16 nM) and produced a 29 mV depolarising shift in Nav1.7 voltage dependence of activation. Cd1a (400 pmol) fully reversed Nav1.7-evoked pain behaviours in mice without producing side effects. In conclusion, Cd1a inhibited two anti-nociceptive targets, appearing to interfere with Cav2.2 inactivation gating, associated with the Cav2.2 α-subunit pore, while altering the activation gating of Nav1.7. Cd1a was inactive at some of the Nav and Cav channels expressed in skeletal and cardiac muscles and nodes of Ranvier, apparently contributing to the lack of side effects at efficacious doses, and suggesting potential as a lead for development of peripheral pain treatments.


Introduction
tested for activity against human (h) Ca v 2.2 channels in SH-SY5Y cells using a FLIPR TETRA (Fluorimetric Imaging Plate Reader, Molecular Devices, California, USA) Ca 2+ imaging assay, as detailed below [21]. Venoms were re-tested at 4 μg/mL to identify the most active venoms. C. darlingi venom produced complete block of Ca v 2.2-mediated Ca 2+ responses at both 4 and 40 μg/well and it was selected for further fractionation to isolate the active peptide.
Automated protein sequencing was performed by the Australian Proteome Research Facility (Sydney, NSW, Australia), using an Applied Biosystems 494 Procise Protein Sequencing System. Briefly, the purified peptide (Cd1a) was dissolved in 25 mM ammonium bicarbonate, pH 8.0 and reduced with DTT at 56˚C for 0.5 h, then alkylated using iodoacetamide at room temperature for 0.5 h. The reduced/alkylated Cd1a was then purified using RP-HPLC (using a Zorbax 300SB-C18 column 3 × 150 mm). A single major peak eluting at 18.3 min was collected, evaporated to 50 μL and loaded onto a Precycled Bioprene-treated disc for Edman sequencing.
Cell culture and fluorimetric assays SH-SY5Y cells were maintained in Roswell Park Memorial Institute (RPMI) medium supplemented with 10% foetal bovine serum (FBS) and 2 mM L-glutamine at 37˚C in a 5% humidified CO 2 incubator. HEK293 cells stably expressing hNa v 1.1-1.8 channels α-subunits coexpressed with the Na v β 1 subunit (Scottish Biomedical, Glasgow, UK; except Na v 1.6, which was generated by GlaxoSmithKline, Stevenage, UK) were maintained at 37˚C in a 5% humidified CO 2 incubator in Dulbecco's Modified Eagle Medium (DMEM; Invitrogen, MA, USA) supplemented with 10% FBS, 2 mM L-glutamine, and kept under selection antibiotics as per manufacturer's protocol. Cells were plated 48 h prior to assay at a density of 30,000-50,000 cells/well on 384-well black-walled imaging plates (Corning, NY, USA). All the cells were used up to maximum 20 passages.
The FLIPR assays were performed based on previously described protocols [17,21,23]. For the FLIPR calcium-based assays, SH-SY5Y cells were pre-incubated for 30 min (37˚C in a humidified 5% CO 2 incubator) with a Calcium 4 fluorescence dye (Molecular Devices) diluted in PSS buffer + 0.1% BSA and nifedipine (10 μM) when testing for hCa v 2.2 activity. Conversely, to test for Ca v 1 activity, cells were pre-incubated with ω-conotoxin CVID (1 μM) for 30 min. After the incubation period the plates were transferred to the FLIPR and camera gain and intensity adjusted for each plate to yield 800-1000 arbitrary fluorescence units (AFU) baseline fluorescence. Ca 2+ responses were measured using a cooled CCD camera with excitation at 470-495 nM and emission at 515-575 nM. Ten baseline fluorescence readings were taken prior to addition of crude venoms, venom fractions or purified peptides diluted in PSS + 0.1% BSA, then fluorescence readings were taken every second for 300 s. After 300 s, activators were added to the cells and fluorescence readings recorded each second for a further 300 s. Endogenous hCa v channels were activated with 90 mM KCl/5 mM CaCl 2 .
The FLIPR membrane potential assays for sodium channels were performed using HEK293 cells expressing hNa v 1.1-1.8. The cells were loaded with Membrane Potential Assay Kit Red (Molecular Devices) reconstituted in PSS + 0.1% BSA and incubated at 37˚C for 30 min before transfer to the FLIPR. After ten baseline fluorescence readings, changes in fluorescence (excitation 510-545 nm; emission 565-625 nm) in response to addition of antagonists (crude venoms or peptides) were measured every second for 300 s. After 300 s, activators (Na v 1.8: deltamethrin at 100 μM, Sigma-Aldrich; Na v 1.6: veratridine at 20 μM, Abcam, Melbourne, VIC, Australia; and all other Na v isoforms: veratridine 70 μM) were then added and the responses monitored for a further 300 s, except for Na v 1.8 cells, where responses were measured for 1800 s.
A four-parameter Hill equation was used to fit concentration-response curves by nonlinear regression analysis (GraphPad Prism v5.0, San Diego, CA, USA). Results are presented as the mean ± standard error of the mean (SEM) of 3-6 replicates on 384-well plates for each independent experiment, performed 3-6 times. Statistical significance was determined using analysis of variance (ANOVA) or a Student's t-test.
[ 125 I]-GVIA binding assay SH-SY5Y cell membranes were prepared using an adaptation of the method of Wagner et al [24]. The cells were harvested using trypsin/ethylenediaminetetraacetic acid (EDTA; Lonza, Basel, Switzerland), washed once with Dulbecco's phosphate-buffered saline (DPBS; Sigma-Aldrich) and centrifuged for 4 min at 500 × g. After centrifugation, the supernatant was discarded, then the pellet was re-suspended in 10 mL binding assay buffer ( 2) containing complete protease inhibitor (Roche Diagnostics, NSW, Australia) and sonicated. The homogenate was then centrifuged at 4˚C for 30 min at 40,000 × g, the supernatant discarded, and the pellet dissolved in binding assay buffer containing 10% glycerol, before storage at -80˚C prior to use. A bicinchoninic acid (BCA) assay kit (Thermo Scientific, Waltham, MA, USA) was used for protein quantification as per manufacturer's instructions.
Radiolabelled conotoxin GVIA ([ 125 I]-GVIA; iodinated at Tyr22) was prepared using IODOGEN as previously described [25], or purchased (Perkin Elmer, USA), and stored at 4˚C for use within 20 days. On the day of the assay, the membranes were thawed on ice and reconstituted to 10 μg/50 μL (mouse) or 10-20 μg/50 μL (SH-SY5Y) in binding assay buffer containing 2% complete protease inhibitor and 0.1% BSA. Stock [ 125 I]-GVIA was diluted to 20,000 cpm/50 μL ([30 pM]). For displacement studies, [ 125 I]-GVIA was incubated with mouse brain membranes or SH-SY5Y membranes and varying concentrations of the competing ligand on 96-well plates. The plates were incubated with shaking for 1 h at room temperature and vacuum filtered through a glass-fibre filter pre-soaked in 0.6% polyethyleneimine (PEI) to reduce non-specific binding and washed with buffer containing 20 mM HEPES and 125 mM NaCl at pH 7.2 using a Tomtec harvester vacuum system (CT, USA). The filters were then dried at 37˚C before being placed in sample bags and soaked in liquid scintillant. Retained radioactivity was then counted using a MicroBeta JET microplate liquid scintillation counter (Wallac, Finland). Non-specific binding was determined in the presence of 50 μL of unlabelled peptides. Specific binding was calculated as the difference between total and non-specific binding. A one-site model was fitted to the data using GraphPad Prism v5.0. Results are presented as the mean ± SEM of 3-6 replicates, performed at minimum 3 independent experiments. Statistical significance was determined using analysis of variance (ANOVA) or a Student's t-test, with statistical significance defined as p < 0.05.

Electrophysiological properties of rCa v 2.2 channels
HEK tsA-201 cells used for the Ca v 2.2 patch-clamp experiments were cultured and transiently transfected with native rat (r) Ca v 2.2 (Ca v α 1b co-expressed with Ca v β 1b and Ca v α 2 δ subunits) or mutant rCa v 2.2-G1326P, as previously described [26]. Transfected cells were incubated for 48 h at 37˚C and 5% CO 2 , re-suspended with 0.25% (w/v) trypsin-EDTA (Invitrogen) and plated onto glass coverslips at least 3-4 h before the patch clamp experiments. Ca v currents were measured by conventional whole-cell patch clamp using an Axopatch 200B amplifier in combination with Clampex 9.2 software (Molecular Devices), as previously described [27]. Cd1a was prepared daily in external solution containing 0.1% BSA and applied to the cells with a gravity-driven micro-perfusion system. The external recording solution for calcium channel recordings contained (in mM): 114 CsCl, 20 BaCl 2 , 1 MgCl 2 , 10 HEPES 10 glucose, adjusted to pH 7.4 with CsOH. For voltage-clamp recordings, 5 μM CdCl 2 was also added to the external solution to inhibit Ca v channels. For all recordings, the internal patch pipette solution contained (in mM): 108 CsMeSO 4 , 2 MgCl 2 , 11 EGTA, 10 HEPES, adjusted to pH 7.4 with CsOH supplemented with 0.6 mM GTP and 2 mM ATP immediately before use. After establishment of the whole-cell configuration, cellular capacitance was minimized using the analog compensation available on the amplifier. Series resistance was < 10 MO and compensated to > 85% in all the experiments. Data were filtered at 1 kHz (8-pole Bessel) and digitized at 10 kHz with a Digidata 1320 interface (Molecular Devices). For current-voltage (I-V) relationship studies, the membrane potential was held at -110 mV and cells were depolarized from -80 to 20 mV in 10-mV increments. For steady-state inactivation studies, the membrane potential was depolarized by test pulses to 0 mV for Ca v 2.2, after 3.6 s conditioning pre-pulses ranging from -110 to 0 mV. Individual sweeps were separated by 12 s.
The cells expressing hNa v 1.7 were maintained with a holding potential of -100 mV and I-V relationships determined using a family of 500 ms conditioning pulses from -120 mV to +70 mV in 5-mV steps, followed by depolarization to 0 mV to assess the voltage dependence of fast inactivation. Each sweep was separated by 20 s to allow complete recovery from inactivation. State-dependence was assessed after a 10 min compound incubation to ensure steady-state inhibition for each concentration. In order to assess compound activity at the partially inactivated/open state, a series of 10 × 50 ms depolarizations to 0 mV were measured after corresponding conditioning pulses to -55 mV for 8 s with a 50 ms recovery before the test pulse, and cycled over a 12 s period for recovery of inactivation. To assess possible interaction with the voltage sensor domain, a triple-pulse protocol was used, comprising two steps to 0 mV for 50 ms separated by a strong depolarization depolarisation step to 200 mV for 50 ms, with 20 ms recovery to the -100 mV holding potential between each step.

Electrophysiology data analysis
The Ca v channel electrophysiology data were analyzed using Clampfit 9.2 (Molecular Devices). Curves were fitted using Origin 7.5 software (Northampton, MA, USA). Electrophysiology data for Na v channels were assessed using the Sophion QPatch Assay Software v5.0, with curves fitted using GraphPad Prism v5.0. I-V relationships were fitted with using a modified Boltzmann equation: where V m is the test potential, V a is the half-activation potential, E rev is the reversal potential, G max is the maximum slope conductance, and k a reflects the slope of the activation curve. The voltage dependence of steady-state inactivation was calculated by dividing the amplitude of the test current (I) by the maximal current elicited (I o ). Steady-state inactivation curves were fitted using the Boltzmann equation: where V h is the half-inactivation potential, k is the slope factor and V m is the holding voltage. Statistical significance was determined by paired or unpaired Student's t-tests and one-way or repeated measures ANOVA for n ! 6 independent experiments, followed by Dunnett's post-hoc test. Differences were considered significant if p < 0.05. Data are expressed as mean ± SEM.

Animal behaviour assessment
All the experiments involving animals were conducted according to the International Association for the Study of Pain Guidelines for the Use of Animals in Research (http://www.iasp-pain.org), in agreement with the Animal Care and Protection Regulation Qld (2012), and the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes, 8th edition (2013) (http://www.nhmrc.gov.au). Ethics approval was obtained from the University of Queensland Institutional Animal Ethics Committee. C57BL/6J, an inbred mouse strain widely used in biomedical research [28], was used in all animal tests. Prior to experimentation, adult C57BL/6J male mice (5-8 weeks of age, average weight 20-25 g) were housed in groups of 2-4 under 12 h light-dark cycle, with free access to standard rodent chow and water. After experiments animals were euthanized by CO 2 asphyxiation. All efforts were made to reduce the number and minimize the suffering of animals.
To test the analgesic potential of Cd1a, while avoiding potential off-target related side effects produced by systemic administration, we used a mouse model of Na v 1.7-mediated pain, as previously described [29], based on intraplantar injection of the α/β scorpion toxin OD1 [30]. Briefly, a concentration of OD1 found to induce pain-like behaviour in mice (300 nM/40 μL diluted in sterile saline/0.1% BSA) [29] was injected subcutaneously into the subplantar surface of the left hind paw of mice (intraplantar, i.pl.) under light isoflurane anaesthesia (3%). Control animals received OD1 (300 nM) and treated animals received Cd1a (0.1 nM-10 μM/40 μL), Cm1a, Cm1b or tetrodotoxin (TTX) (1 μM/40 μL co-injected with OD1). Immediately after injection, the mice were placed into polyvinyl boxes (10 × 10 × 10 cm) and, after recovery from isoflurane, animals were monitored with a video camera placed under the boxes. Spontaneous pain behaviour (paw flinches, shakes and licks) and side-effects were visually assessed over 10-30 min by an observer blinded to all groups. Motor performance was assessed using the Parallel Rod Floor Test, with each peptide administered i.pl. 5 min before the testing. Distance travelled (m) and number of foot slips were recorded using ANY-Maze software (Stoelting Co., version 4.70, Wood Dale, IL, USA). The ataxia index was calculated by dividing the number of foot slips by the distance travelled.
Data were fitted using GraphPad Prism v5.0 and are presented as mean ± SEM (n = 6-12 mice). Statistical significance was determined using analysis of variance (ANOVA) with a Dunnett's post-test, with significance defined as p < 0.05.

Blowfly toxicity assay
To determine Cd1a activity in insects we used a previously reported method [31] with some modifications. Synthetic Cd1a was dissolved in insect saline and injected into the ventro-lateral thoracic region of adult sheep blowflies (Lucilia cuprina) with an average mass between 26.9 and 29.3 mg. A 1.0 mL Terumo Insulin syringe (BD Ultra-Fine, Terumo Medical Corporation, MD, USA) with a fixed 29 G needle fitted to an Arnold hand micro-applicator (Burkard Manufacturing Co. Ltd., England) was used to inject a maximum volume of 2 μL per fly. All flies were individually housed in 2-mL tubes and paralytic activity and lethality were determined at 1 h and 24 h post-injection. A total of three tests were carried out and for each test seven doses of Cd1a (n = 10 flies per dose) and the appropriate control (insect saline; n = 20 flies each) were used. PD 50 values were calculated as described previously [31] using Prism 6.

Materials
TTX, Cm1a and Cm1b were purchased from Alomone Labs, Jerusalem, Israel. OD1 and CVID were synthesized as previously described [30,32,33]. All the other chemicals were purchased from Sigma-Aldrich Australia, unless otherwise indicated.

Discovery, synthesis and biochemical analysis of Cd1a
To find novel Ca v 2.2 inhibitors, we screened 60 spider venoms using fluorimetric assays and SH-SY5Y cells expressing hCa v 2.2. Approximately 30% of the venoms tested inhibited ! 50% of the hCa v 2.2 responses (S1 Table), confirming spider venoms are rich sources of Ca v 2.2 inhibitors. Venom from the African spider C. darlingi (spider shown in Fig 1A) fully inhibited hCa v 2.2 (estimated 37 ng/μL). A search on ArachnoServer, a database listing spider venom toxins [20], indicated that there was no other Ca v 2.2 inhibitor discovered from C. darlingi venom and thus, we selected this venom for fractionation. A single fraction eluting as a sharp peak at~30% solvent B and~99% purity (based on the HPLC profile) inhibited hCa v 2.2 ( Fig  1A). MALDI-TOF MS analysis indicated that the active fraction was dominated by a single mass (charged monoisotopic [M+H] + ) of 4028.2 Da (Fig 1B). Edman sequencing revealed a 33-residue peptide sequence (DCLGWFKSCDPKNDKCCKNYSCSRRDRWCKYDL-NH 2 ), with MALDI-TOF analysis indicating that the peptide was C-terminally amidated (~-1 Da difference from the mass calculated for the free acid form of the Edman-derived sequence).
Cd1a was chemically synthesised using stepwise SPPS. A prominent disulfide bond isomer purified by HPLC to > 97% homogeneity had identical molecular mass to, and co-eluted with, native Cd1a (Fig 1B). Synthetic Cd1a inhibited Ca v 2.2, further confirming that the synthetic peptide was correctly folded.

Sequence homology studies
Surprisingly, BLAST searches on ArachnoServer database [20,34] revealed that Cd1a was poorly similar to other Ca v 2.2 inhibitors. The most similar Ca v inhibitor isolated from a spider venom was ω-TRTX-Hs1a (Huwentoxin-X) [35], which shares only 40% identity with Cd1a. Cd1a lacks most of the functional residues shown to be important for Ca v 2.2 block by ω-conotoxin CVID (Fig 1C), the most selective Ca v 2.2 inhibitor described to date [33].

Cd1a binding on Ca v 2.2 did not completely overlap the GVIA binding site
We used radiolabeled GVIA ( 125 I-GVIA) in competitive binding studies to investigate the Cd1a mode of action. ω-Conotoxin GVIA is known to bind to Ca v 2.2 channel α-subunit pore, in a region localized within the external EF hand motif of domain III S5-S6 (see S1 Fig) [26,39]. Surprisingly, whereas unlabeled GVIA (used as a control) fully displaced radiolabeled 125 I-GVIA from both human SH-SY5Y (IC 50 of 0.18 ± 0.01 nM) and mouse brain cell Novel analgesic Nav1.7/Cav2.2 inhibitor from Ceratogyrus darlingi membranes (IC 50 of 0.27 ± 0.01 nM) (Fig 2B) with potencies similar to previous described [21,24], Cd1a (10 μM) did not displace 125 I-GVIA from these membranes, indicating that the Cd1a binding site does not overlap the GVIA binding site on Ca v 2.2 channels.

Cd1a did not act as a classical gating modifier toxin on Ca v 2.2 channels
The Cd1a effect was investigated on Ba 2+ currents measured using whole-cell voltage-clamp electrophysiology on rCa v 2.2 (Ca v α 1B , Ca v β 1b + Ca v α 2 δ) expressed in tsA cells. Cd1a (3 μM) inhibited the currents by~50% (n = 9) and there was no recovery from block during a 5 min wash-off period (Fig 3A). The reversal potential and half-activation voltage for Cd1a-treated cells (E rev = +51.2 mV, V a = -5.3 mV) were not significantly different from untreated cells (E rev = +52.4 mV, V a = -1.7 mV) ( Fig 3B). In contrast, Cd1a produced a statistically significant hyperpolarizing shift in the half-inactivation potential of~8 mV (V h control = -46 mV; V h Cd1a = -54.6 mV (Fig 3C). The absence of a measurable effect on the voltage dependence of activation under our conditions indicates. Cd1a has a unique mode of action at Ca v 2.2 channels, differing from that of classical gating modifier toxins, such as ω-grammotoxin SIA and ω-IVA [39].

Cd1a affected mutant rCa v 2.2-G1326P function
We tested the ability of Cd1a to inhibit rCa v 2.2 carrying a G1326P mutation (Ca v 2.2-G1326P), which lies on the large extracellular loop region of the Ca v 2.2 α subunit. This mutation results in localized structural disruption within the domain III S5-H5 region associated with the external EF hand motif within domain III S5-S6 [26, 40,41] that alters ω-conotoxin GVIA and MVIIA affinity and reversibility [41], associating this location with the binding site for these toxins. Whereas wild-type rCa v 2.2 channels showed no recovery from Cd1a-evoked block, Cd1a inhibition of the mutant Ca v 2.2-G1326P channel was reversed following a 5 min washout (Fig 3A). In addition, Cd1a induced a significant shift in E rev of the mutant channel (control = 61.0 mV, Cd1a = 52.5 mV) in contrast to its lack of effect at the wild-type channel. Novel analgesic Nav1.7/Cav2.2 inhibitor from Ceratogyrus darlingi The voltage dependence of activation of the mutant, like the native rCa v 2.2, was unaffected (V a control = 7.0 mV, V a Cd1a = 7.2 mV), whereas the voltage dependence of inactivation underwent a small but significant shift (V h control = -28.9 mV; V h Cd1a -38 mV (Fig 3D and 3E). The combined data suggest that, despite no overlap in binding sites demonstrated by the inability of Cd1a to displace GVIA in biochemical studies, there is some overlap in the molecular determinants of channel interactions between these two toxins. However, more extensive channel mutagenesis is required to fully define the Cd1a binding site.

Cd1a activity on 8 heterologously expressed hNa v channels
Due to the high similarity between Cd1a and the Na v channel inhibitors Cm1a and Cm1b, we assessed the effect of Cd1a on hNa v 1.1-1.8 channels heterologously expressed in HEK293 cells using fluorimetric assays, and compared with the effects of Cm1a and Cm1b using the same assays. Overall, Cd1a activity was not statistically different to Cm1a and Cm1b (two-way Anova), with each peptide inhibiting Na v 1.1, Na v 1.2 and Na v 1.7 (IC 50 values in Fig 4A,  Table 1). Cd1a (30 μM) and Cm1a (10 μM) (Fig 4B) were inactive at Na v 1.3-Na v 1.6 ( Fig 4A-4C, Table 1), whereas Cm1b (Fig 4C) was also inactive at Na v 1.4-Na v 1.5 (up to 10 μM), but inhibited hNa v 1.3 and Na v 1.6. Cd1a was the only peptide with activity at Na v 1.8 ( Fig 4A,  Table 1) in our assays, in contrast with previous report on rat clones using patch clamp electrophysiology [36], where 2 μM Cm1a and Cm1b inhibited 55 and 40% of Na v 1.8 current, respectively.
The Cd1a mode of action on hNa v 1.7 We assessed the Cd1a effect on hNa v 1.7 channels stably expressed in HEK293 cells using planar patch-clamp electrophysiology. Cd1a inhibited the hNa v 1.7 peak current in a concentration-dependent manner, with an IC 50 in the low nanomolar range (16.0 ± 3.0 nM, n = 10; Fig  5A). An I-V family of conditioning pulses and their corresponding G-V curves were used to determine the effect of Cd1a on the voltage dependence of activation (V a ), while depolarization to 0 mV following a conditioning pulse was used to assess its effects on the steady-state inactivation (V h ) (Fig 5B). Interestingly, Cd1a (100 nM) induced a significant 29 mV depolarizing shift in V a (control -20.4 ± 0.3 mV; Cd1a 8.6 ± 1.0 mV, n = 11) (Fig 5C). Furthermore, at 100 nM Cd1a induced a small 3.8 mV hyperpolarizing shift of the voltage dependence of inactivation (V h ) of hNa v 1.7 (control -58.9 ± 0.4 mV; Cd1a -62.7 ± 0.5 mV, n = 9) (Fig 5D).
We used a triple pulse protocol previously described [36,42] to identify if Cd1a interacted with the Na v 1.7 voltage sensor domains. Under control conditions there was no significant change between a first (P1) and third (P3) pulses separated by a strong positive pulse (P2) (Insert Fig 5E). Cd1a at 1 μM fully inhibited P1 as expected, however, the P3 current was not fully inhibited, indicating partial relief of block by P2. This suggests that Cd1a interacts with one or more voltage sensor domain of hNa v 1.7 channels.
Cd1a is analgesic in a Na v 1.7 mouse model of peripheral pain Intraplantar injection of the scorpion toxin OD1 evokes spontaneous pain behaviours (paw licks, shakes and flinches) in mice (Fig 6A), an effect mostly mediated by Na v 1.7 channels [29]. Intraplantar injections of Cd1a (400 pmol) completely reversed OD1-evoked spontaneous pain behaviours for at least 30 min (Fig 6A). The Cd1a effect was concentration dependent (IC 50 = 0.36 ± 0.12 pmol) (Fig 6B). For comparison, we tested Cm1a, Cm1b and TTX in the same mouse model. At a dose that all peptides with activity at Na V 1.7 reversed spontaneous pain behaviours (40 pmol), TTX caused sedation, and reduced motor coordination Data are from the FLIPR fluorimetric membrane potential assays. Cd1a and Cm1a-b inhibited a range of Na v isoforms with variable potency (see IC 50 values for the three toxins on Na v 1.1-Na v 1.8 in Table 1). Data points are mean ± S.E.M (n = 3-6 replicates).
( Fig 6C and 6D) in all mice, whereas Cd1a, Cm1a and Cm1b did not produce any apparent side effects.

Cd1a reversibly paralysed sheep blowfly
Cd1a produced paralytic effects in adult sheep blowfly, with a PD 50 measured 1 hour after injection of 1318 ± 58 pmol/g (Fig 7). Remarkably, however, all flies fully recovered from the initial paralytic effects induced by Cd1a at doses up to 8.17 nmol/g.

Discussion
We describe the discovery and characterization of Cd1a, the first Na v /Ca v inhibitor peptide reported from the African theraphosid spider C. darlingi. Cd1a belongs to NaSpTx family 1 [19], a class of promiscuous toxins that can modulate a range of ion channels, including Na v , Ca v , K v , mechanosensitive and proton-gated ion channels. Peptides from NaSpTx family 1 share an ICK structural motif that typically provides resistance to heat denaturation and proteolysis [37,44], features that are potentially advantageous for drug development. Interestingly, Cd1a inhibits important peripheral nociceptive targets, including Na v 1.7, Na v 1.8 and Ca v 2.2, but not cardiac Ca v and Na v channels (Ca v 1.3, Ca v 3.1 and Na v 1.5), or Na v channels found in skeletal muscle (Na v 1.4) or nodes of Ranvier in motor nerves (Na v 1.6). The Cd1a selectivity profile is consistent with its analgesic efficacy and lack of side effects at maximal efficacious doses in a mouse model of peripheral pain shown in this work.
Cd1a has little sequence similarity to other venom peptides that inhibit Ca v channels, likely reflecting its relatively low potency against Ca v 2.2 (IC 50 of~3 μM at rat and human Ca v 2.2). Like Cd1a, the related peptides Cm1a and Cm1b from C. marshalli [36] are known Na v inhibitors and have little sequence similarity to Ca v inhibitors. However we found for the first time that these peptides inhibit hCa v 2.2 with moderate potency, but were inactive at hCa v 1.3 and hCa v 3.1. Our results contrasted with a previous report on patch clamp electrophysiology, which indicated that Cm1a and Cm1b were inactive at HVA calcium currents of sensory neurons [36]. However Cm1a and Cm1b were tested at a maximum concentration of 100 nM [36], thus at a higher concentration it's conceivable that these peptides would also be active in their system. Unlike classical Ca v channel gating modifier toxins, Cd1a didn't affect activation gating of rCa v 2.2 but instead induced a small shift in the voltage-dependence of inactivation, suggesting that at this channel Cd1a doesn't act as a typical gating modifier toxin. Cd1a activity was enhanced at the rCa v 2.2α1B-G1326P mutant channel while channel inhibition became partially reversible, indicating possible partial overlap with the binding site for ω-conotoxins.  In addition, Cd1a caused a small change in reversal potential of the mutant channel but not the wild-type, while channel inhibition became partially reversible, indicating binding near the ion permeation/selectivity pathway. However, competitive binding studies revealed that the Cd1a binding does not completely overlap with the GVIA binding site on mutant Ca v 2.2. Thus, we suggest a possible allosteric coupling between the Cd1a and GVIA binding site that is introduced by the rCa v 2.2α1B-G1326P mutation. Residues in the outer vestibule of the channel pore may allosterically affect Cd1a binding and its ability to interact with the inactivated state of the channel. Precisely how Cd1a influences the voltage dependence of inactivation Cd1a (1 μM) fully inhibited peak current in P1. Current inhibition was partially reversed in P3 using the positive depolarizing pulse protocol suggesting interaction of Cd1a with one of the Na v 1.7 voltage sensors. Data points are expressed as mean ± S.E.M (n = 9-13 replicates). Novel analgesic Nav1.7/Cav2.2 inhibitor from Ceratogyrus darlingi remains to be determined, but it is possible that the toxin interacts with one or more of the S6 regions, which have been linked to fast inactivation [45]. The high level of homology between Cd1a and the Na v inhibitor peptides Cm1a and Cm1b suggested that Na v channels may be the primary high-affinity target for Cd1a. Indeed, in electrophysiological studies Cd1a was~200-fold more potent at recombinant hNa v 1.7 than rCa v 2.2, inhibiting hNa v 1.7 with an IC 50 of~15 nM. Whereas we have previously shown that species differences can lead to discrepancies in potencies [21], the IC 50 for Cd1a at human and rat Ca v 2.2 were similar in both in patch clamp and fluorescence assays (~3 μM), confirming the primary target for Cd1a is Na v 1.7.
Using a FLIPR membrane potential dye assay, we identified that Cd1a, Cm1a and Cm1b were similarly active across a range of hNa v channels, including Na v 1.1-1.2 and Na v 1.7. However, Cd1a potency at Na v 1.7 was lower using the FLIPR assays than when using patch clamp electrophysiology (~200-fold), consistent with previous reports [46]. In addition we could not detect Na v 1.8 activity for Cm1a and Cm1b, in contrast with previous report [36]. Although an explanation for these differences remains to be elucidated, a number of factors may affect between-assay peptide potency. These include different expression systems and β subunit combinations, species differences, influence of membrane potential dye, and the requirement for channel activation by veratridine in fluorescence assays. Nonetheless, fluorescence-based assays allow high-throughput screening and rapid assessment of pharmacology. While the rank order of potency across Na v isoforms is typically conserved, differences in potencies between patch clamp electrophysiology ('the gold standard method for assessing ion-channel function') and fluorescence based assays have been described [46,47,48].
Although Cd1a didn't affect the steady-state voltage parameters of activation of rat Ca v 2.2, the human Na v 1.7 inhibition was driven by a depolarising shift in V a . This effect on Na v 1.7 is consistent with gating modifying activity through one or more of the voltage sensor domains Novel analgesic Nav1.7/Cav2.2 inhibitor from Ceratogyrus darlingi (VSD) [36,39,42]. A forced outward movement of one or more Na v voltage sensor domains has been achieved with strong positive depolarizing pulses, releasing gating modifier toxins from the voltage sensor domains [36,42] and leading to recovery from toxin inhibition. In support, the strong positive-depolarising pulse we used in our patch clamp experiments enhanced recovery from Cd1a inhibition by~70% and inhibition of inward but not outward current were observed, similar to other gating modifier toxins [36,42].
Due to its pharmacology profiling and mode of action, with the highest activity at Na v channels, Cd1a was named β-theraphotoxin-Cd1a (or β-TRTX-Cd1a as abbreviation), based on the rational nomenclature devised for spider toxins [49]. While the exact site of Cd1a interaction with Na v channels remains to be elucidated, spider-venom peptides with similar modes of action have been found to bind to VSDs DI-DIII but not to DIV, which appears to influence inactivation rather than activation [50][51][52].
We assessed the analgesic potential of Cd1a by examining its ability to reverse spontaneous pain induced by intraplantar injection of the Na v 1.7 channel activator OD1 [29,30,43]. Remarkably, Cd1a was able to completely reverse the OD1-induced nociceptive behaviour in mice, with no apparent off-target related side effects at the highest dose tested and no motor side effects. These results contrasted with TTX, which also fully reversed the OD1-induced pain behaviour at the same dose, but caused sedation and motor incoordination, suggesting off-target neuronal and/or skeletal muscle effects associated with its inhibition of Na v 1.1-1.4 and/or Na v 1.6 [53]. Consistent with the lack of side effects seen for Cd1a in mice, Cd1a was inactive on cardiac and peripherally expressed Ca v 1.3, Ca v 2.1 Na v 1.4, Na v 1.5 and Na v 1.6 (30 μM). Cd1a was potent at centrally expressed Na v 1.1-1.2, however, it is unlikely to cross the blood-brain barrier and interact with these channels centrally [9].
In future, Cd1a needs to be tested in other more conventional animal models of pain, such as the inflammation and neuropathic pain models, to confirm its usefulness as a peripheral analgesic lead, and to provide information tools to compare Cd1a with peptides from other animal venoms.
Whereas the analgesic effect of Cd1a might be useful as a therapeutic lead, we wondered about the ecological significance of Cd1a. Pain-inducing venom components for defensive purposes have been previously reported from tarantula venoms [31], but the analgesic activity of Cd1a does neither endow the spiders with an evolutionary advantage nor support a potential defensive purpose. Hence, we assumed that it might be used for predation and with insects and other invertebrates being the main prey of tarantulas [54] we tested Cd1a in a blowfly toxicity assay. The fact that Cd1a induced paralysis in blowflies supports a potential role for Cd1a in predation.
In summary, we have isolated and pharmacologically characterized Cd1a, a novel peptide inhibitor that showed insecticidal effects and acts on the anti-nociceptive targets Ca v 2.2, Na v 1.7 and Na v 1.8. Cd1a additionally inhibits the central Na v 1.1-1.2, but not peripheral offtarget channels such as Ca v 1.3 and Ca v 2.1 or Na v 1.4-1.6. Interestingly, Cd1a acts at Ca v 2.2 and Na v 1.7 with distinct modes of action, inhibiting Ca v 2.2 activity affecting near the pore region, but not overlapping the ω-conotoxin binding site, interfering with inactivation with no apparent effects on the activation gating of the channel. Conversely, at Na v 1.7 Cd1a acts as a typical gating modifier toxin, interacting with one or more of the voltage sensor domains and changing the gating properties of this channel. The primary structure of Cd1a strongly suggests that it will fold into an ICK motif that is expected to provide a high level of chemical, thermal and biological stability. Thus, Cd1a may be a useful lead for development of a peripherally acting analgesic. In future studies it will be important to examine Cd1a analgesic potential in a wider range of pain models to confirm its usefulness as an analgesic lead.
Supporting information S1